LiveKit Voice AI Agent with Function Calling & Knowledge Base (Ubuntu + Web Client)

We want to test a voice AI agent our team can talk to from the browser, for example on the phone in the car for a quick morning brief with data from our systems. We’ll start lean with a working prototype we can show the team, and it may grow into a bigger project later. What we need: * LiveKit Agents running on our Ubuntu server, connected to LiveKit Cloud, and still running after a server restart * A voice pipeline built on hosted services: speech-to-text, a fast LLM, and text-to-speech. Your recommendation is welcome. * A simple, mobile-friendly web page to start a voice call with the agent * A small knowledge base the agent can answer from * Function calling with read-only access to our internal API to retrieve data * A webhook connection to our internal Claude setup: the agent sends a request, Claude runs one predefined read-only skill and sends the result back. While it waits, the agent tells the user it’s working on it. * Short notes on how to change the prompt, voice, knowledge base and tools, and how to restart the agent You should have: * Built LiveKit voice agents before (please share an example) * Experience with function calling and webhooks via VoiceAgents/LiveKit * Solid Linux/server experience In your proposal, tell us: What LiveKit projects you’ve done